GPU Cloud Provider · Redwood Shores, California, USA
Oracle Cloud
Oracle Cloud Infrastructure (OCI) provides advanced AI infrastructure designed for a wide range of AI workloads, including model training and inference. OCI supports scale-out capabilities up to zettascale with bare metal instances, equipped with the latest NVIDIA GPUs and high-bandwidth, low-latency RDMA networking.
GPUs
1
Founded
Oracle Corporation founded in 1977, with Oracle Cloud developed later.
Countries
21
Data Centers
28
Uptime SLA
99.9%
Team Size
10,000+
GPU Marketplace
NVIDIA A100 80GB SXMOn-Demand
Company Profile
Company TypeHyperscaler
Provider TypeHyperscaler
FoundedOracle Corporation founded in 1977, with Oracle Cloud developed later.
HeadquartersRedwood Shores, California, USA
Legal EntityOracle Corporation
FundingPublic (ORCL)
Team Size10,000+
Infrastructure
GPU FleetNVIDIA H100 SXM5 80GB, NVIDIA H100 NVL, NVIDIA A100 80GB SXM, NVIDIA A10, NVIDIA V100, NVIDIA L40S, AMD MI300X
Network FabricRDMA over Converged Ethernet (RoCE v2)
ConnectivityUp to 3,200 Gbps of cluster network bandwidth, up to 400 Gbps front-end network bandwidth
StorageNVMe SSD, Oracle-managed file storage with Lustre, Block storage, Object storage
Data Center TierTier 3+ equivalent; Oracle-owned and operated data centers with redundant power and cooling
Bare MetalYes — bare metal GPU instances available (BM.GPU series)
AvailabilityGA
EnterpriseGovernmentResearchStartupFinancial Services
Compute & Deployment
On-DemandYes
Spot / InterruptibleYes (preemptible instances available at reduced cost)
Reserved InstancesYes (1-year and 3-year reserved capacity commitments)
Bare MetalYes (bare metal GPU instances including BM.GPU4.8 with A100s and BM.GPU.H100.8 with H100s)
VM-BasedYes (VM-based GPU shapes available alongside bare metal)
Container-BasedYes (Docker and container workloads supported via OCI Container Engine)
KubernetesYes (managed K8s via Oracle Container Engine for Kubernetes, OKE)
Serverless GPUNo
Spin-Up Time2-5 minutes for VMs; bare metal provisioning can take 10-20 minutes
TerraformYes (official provider on HashiCorp registry)
GPU Hardware
Latest GenH100 SXM, H200 SXM, L40S, A100 SXM
Legacy SupportV100, P100
Multi-GPU NodesYes (up to 8x per node)
Max GPUs/Node8
NVLinkYes (NVLink on SXM nodes)
InfiniBandYes (HDR 100Gbps RDMA over RoCE, with RDMA networking via cluster networking)
PCIe vs SXMBoth PCIe and SXM
HGX PlatformYes (HGX H100 8-GPU)
Pricing Model
Per HourYes (primary billing unit)
Per MinuteNo
SubscriptionYes (monthly/annual Universal Credits and Bring Your Own License plans)
Reserved DiscountUp to 70% off with 1-year or 3-year reserved capacity commitments
Spot DiscountNo spot pricing (preemptible instances not widely available for GPU shapes)
Public PricingPartial (some GPUs listed; bare metal and cluster shapes require sales contact)
Hidden FeesLoad balancer charges, public IP address fees, DNS zones billed separately; Universal Credits may obscure per-resource costs
Egress Charges10 TB/month free, then $0.0085/GB (varies by region)
Setup FeesNone for standard accounts; enterprise onboarding may involve professional services fees
Pay-as-you-goYes (Pay As You Go tier available)
Credit SystemYes (Universal Credits prepaid model with committed spend discounts)
Performance & Scaling
Multi-Node TrainingYes (multi-node distributed training with NCCL, RoCE, and InfiniBand support across GPU clusters)
Max Cluster Size4096+ GPUs (via HPC clusters with BM.GPU4.8 and BM.GPU.H100.8 bare metal shapes)
Elastic ScalingManual only (cluster network resizing requires reprovisioning; no live elastic scaling)
Auto ScalingYes (policy-based auto-scaling for VM GPU shapes; limited for bare metal HPC clusters)
InfiniBandYes (HDR 100Gbps InfiniBand on HPC cluster networks for bare metal GPU nodes)
NVSwitchYes (on SXM-based H100 bare metal shapes BM.GPU.H100.8)
SLA99.9%
Perf IsolationYes (dedicated bare metal GPU instances with no hypervisor overhead on HPC shapes)
Noisy NeighborYes (bare metal, no sharing on BM GPU shapes; VM shapes subject to multi-tenant risks)
Developer Experience
OnboardingSelf-service sign-up with free trial; instances can be deployed via web console in minutes; enterprise onboarding with solution architect support available
FrameworksOCI Data Science
SDK LanguagesPython, Java, Go, Ruby, .NET, TypeScript, JavaScript
CLI ToolingFull OCI CLI with instance management, file sync, and automation support; also supports Terraform provider
JupyterVia OCI Data Science service with managed JupyterLab notebooks; also accessible via SSH on GPU instances
TemplatesPyTorch Training, TensorFlow, LLM Fine-tuning, Data Science Notebook Sessions, NVIDIA AI Enterprise stack
Model MarketplaceOCI Marketplace with pre-built GPU images; NVIDIA AI Enterprise integration; no dedicated LLM model hub
DocumentationComprehensive docs with architecture guides, API reference, CLI references, and tutorials; OCI documentation portal is well-structured
API FeaturesCLI, SDK, REST API
Security & Compliance
SecurityComplies with industry-standard security and compliance protocols suitable for government and dedicated clouds.
ComplianceData residency, security, and compliance capabilities are supported, particularly for government and sovereign cloud services.
Public company (NYSE: ORCL) with $50B+ annual revenueFedRAMP authorizedISO 27001, SOC 1, SOC 2, SOC 3 certifiedNVIDIA DGX-Ready Cloud partnerStrategic partnership with NVIDIA for H100 and GB200 capacityDoD IL5 authorized
Data Center Locations
Coverage
CountriesUnited States, United Kingdom, Germany, Netherlands, Switzerland, France, Sweden, Spain, Italy, Canada, Brazil, Chile, India, Japan, South Korea, Australia, Singapore, United Arab Emirates, Saudi Arabia, South Africa, Israel
CitiesAshburn VA, Phoenix AZ, San Jose CA, Chicago IL, London, Frankfurt, Amsterdam, Zurich, Paris, Stockholm, Madrid, Milan, Montreal, Toronto, Sao Paulo, Santiago, Mumbai, Hyderabad, Tokyo, Osaka, Seoul, Sydney, Melbourne, Singapore, Dubai, Jeddah, Johannesburg, Tel Aviv
Multi-Region FailoverYes (automatic and manual failover with cross-region replication)
Latency TiersStandard cloud latency; ultra-low available within same AD cluster (<1ms intra-datacenter)
North AmericaEuropeAsia-PacificMiddle EastLatin AmericaAfrica
Compliance Regions
EU Data ResidencyYes (Frankfurt, Amsterdam, Paris, Stockholm, Madrid, Milan, Zurich)
US Gov CloudYes (FedRAMP authorized, OC2 and OC3 government regions)
India RegionYes (Mumbai, Hyderabad)
Datacenter Locations
Key Strengths
Among the lowest list pricing for H100 bare metal in major hyperscaler category
Second-generation cloud architecture with non-oversubscribed network and consistent performance
Bare metal GPU clusters with RDMA networking (RoCEv2) for low-latency distributed training
Strong enterprise and database integration (Oracle DB, Autonomous DB) alongside AI workloads
OCI Supercluster supports up to 32,768 H100 GPUs with 1.6 Tbps RDMA fabric
Known Limitations
Smaller global footprint than AWS or Azure in some regions
Ecosystem and third-party integrations less mature than AWS/GCP
OCI console and UX can feel complex compared to hyperscaler competitors
Community and developer mindshare significantly smaller than AWS or Google Cloud
Spot/preemptible GPU availability can be limited in certain regions
Additional Information
Support Options
Dedicated support, online resources
Community
Oracle Cloud Community forums; Oracle Developer YouTube channel; Oracle CloudWorld annual conference; limited Discord/Slack presence compared to startup-focused cloud providers
Green Energy
Committed to 100% renewable energy by 2025; climate action report published annually
PUE Rating
1.2
Core Proposition
Oracle Cloud Infrastructure offers high-performance bare metal and VM GPU instances with a high-bandwidth, low-latency flat network architecture optimized for AI/ML workloads, backed by Oracle's enterprise database and application ecosystem.
Notable Customers
NVIDIA
Zoom
Uber
AT&T
Vodafone
8x8
Payment Methods
Credit CardWire TransferOracle Universal CreditsInvoice/PO (Enterprise)OCI Marketplace (AWS/Azure channels)
Last updated March 2026. Information subject to change.